I came late to that party..... but quickly learned that the jFET control used by HP in their -339A can achieve much lower distortion than original..... limiting factor, was as you said, was noise from the osc. but also a smaller contribution by the agc circuit. With lower noise parts (but same jFET) the THD+N can be made much lower..... I went from -95dB down to -105dB or better THD+N with the old 339A model with modern opamp upgrade and some variable trim control in place of fixed R's. I think that THD+N is comparable to more modern 'sound cards'. With the use of notch filters and FFT on residual THD+N, you can get very very low in THD. Below that... the cost of the next 10-20 db's goes very high really fast.
THx-RNMarsh
Last edited:
Thought experiment
Just that, thinking out loud. Say you take the two op-amp Wein circuit and tune it just below oscillation and punt the AGC and level detector and use it as a high Q amplifier/filter. The injection lock now becomes the input, a quick sim shows you can get ~40dB rejection which would bring a good sound card DAC down to the -140dB level. The caps, resistors, and op-amp remain as they always have in terms of distortion contribution.
This still is a one frequency solution.
Just that, thinking out loud. Say you take the two op-amp Wein circuit and tune it just below oscillation and punt the AGC and level detector and use it as a high Q amplifier/filter. The injection lock now becomes the input, a quick sim shows you can get ~40dB rejection which would bring a good sound card DAC down to the -140dB level. The caps, resistors, and op-amp remain as they always have in terms of distortion contribution.
This still is a one frequency solution.
one frequency?
build the best oscillator you can using op-amps and a non-linear element light bulb/ FET whatever, the level changes what only a couple of dB across all the bands. Then add the agc externally to normalize across the bands also use a number of detectors to handle the attenuation function. AFAIK each detector section is good for 15 dB or so. it's expected as the main level is attenuated it goes down toward the noise / spur floor. normal building block approach, all these parts are out there in the wild. I think I used AD part for this at 70MHz modem IF stage once.
build the best oscillator you can using op-amps and a non-linear element light bulb/ FET whatever, the level changes what only a couple of dB across all the bands. Then add the agc externally to normalize across the bands also use a number of detectors to handle the attenuation function. AFAIK each detector section is good for 15 dB or so. it's expected as the main level is attenuated it goes down toward the noise / spur floor. normal building block approach, all these parts are out there in the wild. I think I used AD part for this at 70MHz modem IF stage once.
Last edited:
These are mostly concepts one uses switched cap filters which are lucky to do -60dB THD. Sorry if I wasn't clear I simply want a 1KHz reference tone at an exact frequency, the Christmas bulb oscillator works fine down to -105dB or so. I guess I will just have to try it.
Last edited:
SW --- maybe something here:
https://scholar.google.com/scholar?...ved=0ahUKEwjHrYf6vuTJAhUHxGMKHdzZAWEQgQMIHDAA
THx-RNMarsh
https://scholar.google.com/scholar?...ved=0ahUKEwjHrYf6vuTJAhUHxGMKHdzZAWEQgQMIHDAA
THx-RNMarsh
Last edited:
Unless we can auto tune the high Q amplifier/filter to the injection. Without raising the distortion of course.
The Shibasoku 725x does this with a frequency counter and a lot of very expensive relays.
It does work. I tried this with a variable state and the 339A oscillator back when you did your nosollator. I guess it's a few years now.
Actually the Boonton 1120 works in a pretty similar way with a state variable filter for the analyzer that is tunes withing a few percent and a narrow range nulling circuit. Similar for the oscillator. Using those as interfaces to an ADC/DAC and eliminating the nulling/tuning/agc might very well do what Scott proposes. However not a simple job.
Otherwise the Boonton's oscillator is tunable to a crystal controlled .01 Hz with the flatness of an AC calibrator. Its really great for a number of tasks.
For my specific problem I will look at the THAT (about the worst possible name for a company) part since it would be the easiest to retrofit over the existing Motorola analog multiplier.
Otherwise the Boonton's oscillator is tunable to a crystal controlled .01 Hz with the flatness of an AC calibrator. Its really great for a number of tasks.
For my specific problem I will look at the THAT (about the worst possible name for a company) part since it would be the easiest to retrofit over the existing Motorola analog multiplier.
Exactly right. That's why I chose one for my THD analyzer for the auto-tune circuits, which use detectors fed with I and Q.
Cheers,
Bob
That would work, does it have less sensitivity to op-amp distortion? I don't need I and Q signals but for a quadrature oscillator that is interesting.
BTW I realize that this thing might tweek up like the tickler on a super regen and I don't want to try and make a tunable oscillator, one frequency only locked to a soundcard's output so I can use un-windowed FFT's.
Aren't we repeating what's been discussed before (see http://www.diyaudio.com/forums/equipment-tools/205304-low-distortion-audio-range-oscillator-247.html#post3465620), or did I miss a bit?
Samuel
Samuel
Yes, sorry the memory has faded because I never had the time to build one.
It has added filter slope at the low pass output. 2nd order so you get 6dB more.
The SV can have exceptionally high Q. Up to 200. about 60dB gain at resonance. Easily manually fine tuned with one trim. For a fixed tuned simple. Use the LT1468 and some low noise low disto op amp at the summing / high pass stage. I stuck a AD797 in this position and it worked fine even at unity. The summing amp is inverted so a good place to inject.
You'll need to reduce the gain of the injection to about 30mV or less depending on the frequency. So loads of decoupling there. Probably a 5 meg resistor at the input node.
Even in THIS Forum (with tens of thousands of members) I bet you don't have to take your socks off to count the number of people who fully appreciate this analogy from personal experience.
Dale
